https://scholars.lib.ntu.edu.tw/handle/123456789/632483
標題: | Non-Autoregressive Mandarin-English Code-Switching Speech Recognition | 作者: | Chuang S.-P Chang H.-J Huang S.-F HUNG-YI LEE |
關鍵字: | code-switching; end-to-end speech recognition; non-autoregressive | 公開日期: | 2021 | 起(迄)頁: | 465-472 | 來源出版物: | 2021 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2021 - Proceedings | 摘要: | Mandarin-English code-switching (CS) is frequently used among East and Southeast Asian people. However, the intra-sentence language switching of the two very different languages makes recognizing CS speech challenging. Meanwhile, the recent successful non-autoregressive (NAR) ASR models remove the need for left-to-right beam decoding in autoregressive (AR) models and achieved outstanding performance and fast inference speed, but it has not been applied to Mandarin-English CS speech recognition. This paper takes advantage of the Mask-CTC NAR ASR framework to tackle the CS speech recognition issue. We further propose to change the Mandarin output target of the encoder to Pinyin for faster encoder training and introduce the Pinyin-to-Mandarin decoder to learn contextualized information. Moreover, we use word embedding label smoothing to regularize the decoder with contextualized information and projection matrix regularization to bridge that gap between the encoder and decoder. We evaluate these methods on the SEAME corpus and achieved exciting results. © 2021 IEEE. |
URI: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85117744914&doi=10.1109%2fASRU51503.2021.9688174&partnerID=40&md5=1ad41567fb8af25a7280258fe28a06d2 https://scholars.lib.ntu.edu.tw/handle/123456789/632483 |
DOI: | 10.1109/ASRU51503.2021.9688174 | SDG/關鍵字: | Decoding; Signal encoding; Speech; Speech communication; Switching; Auto-regressive; Autoregressive modelling; Code-switching; Embeddings; End to end; End-to-end speech recognition; Fast inference; Learn+; Non-autoregressive; Performance; Speech recognition |
顯示於: | 電機工程學系 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。